Reservoir Computing with Neural Organoids for Comparative Studies of Neurological Health

Computational capabilities of the brain are not well understood, which makes understanding the pathophysiology of neurological disorders challenging. With advances in cell culture techniques, neural organoids can now be generated with increasing biofidelity, making them ideal for studying biological computation in both healthy and diseased states. Healthy neural organoids have already demonstrated potential to act as reservoir computers, a type of recurrent neural network for solving machine learning tasks. Therefore, a comparative analysis of the computational abilities of organoids from control subjects and subjects with SYNGAP1-related disorder was conducted. All organoids responded to electrical stimulation via a high-density microelectrode array, but SYNGAP1 organoids exhibited larger deviations from the expected interspike intervals in comparison to the delays between subsequent stimulation pulses. Despite this difference, both control and SYNGAP1 organoids performed similarly when acting as reservoirs to predict nonlinear chaotic systems (Mackey-Glass and 1D Hénon). This comparative analysis establishes a preliminary reservoir computing framework, which could be expanded towards standardizing the evaluation of in vitro/in silico combinatory models for investigating neural pathophysiology.